[e2e] Can feedback be generated more fast in ECN?

Panos GEVROS P.Gevros at cs.ucl.ac.uk
Wed Feb 14 05:58:28 PST 2001


imo it depends on which problem one's trying to solve, assuming cooperating 
endpoints the main problem with end-to-end adaptation in principle is lack of 
information (and latency),
there should be no doubt as to which method can provide more precise (and timely) information, 

message processing overhead, seems to be the strongest argument against SrcQuench, so SQ shouldn't be dismissed light-heartedly - i dare say that ``in principle SQ is better than ECN'', 
it is just that the issues around it have not been sufficiently addressed since the overhead seemed prohibitive and its traffic implications probably exaggerated 

of course the generation of SQ messages will have to be regulated in some intelligent fashion, and this would partly address the overhead concerns ; e.g when an estimate of the number of active flows is available then SQ messages could be stochastically distributed across all flows etc.

cheers
Panos






More information about the end2end-interest mailing list